Please enable JavaScript.
Coggle requires JavaScript to display documents.
Large and Fast: Exploiting Memory Hierarchy (Cache (writing (write…
Large and Fast:
Exploiting Memory Hierarchy
Memory Technology
SRAM
fixed access time
DRAM
store charge on a capacitor
Flash
erasable programmable read-only
Magnetic disk
collection of platter
Cache
a safe place for hiding or storing things
directed-mapped cache
memory location is mapped to one cache
(Block address) modulo (# blocks in cache)
valid bit
associated block with valid data
tag
address infomation to identify
cache missing
request for data not in the cache
PC to memory -> read & wait ->
write cache & data -> restart
writing
write-through
always update cache, memory
write-buffer
queue while data is waiting to be written
write-back
only update cache & modify later
Virtual Memory
use main memory as a cache for secondary storage
address
physical memory
address of main memory
virtual address
memory in a virtual space
page fault
accessed page is not present in main memory
address translation
mapping from virtual address to memory
TLB
making address tranlation fast
translation-lookaside buffer
cache that tracks of recently used address mappings
Virtual Machine
managing software
abstraction that can run complete software stack
managing hardware
allows separate software stack to run independently
Virtual Machine Monitor
present software interface to guest software
behave as if running on native hardware
cannot change allocation of resource directly
Dependable Memory Hierarchy
Failure
service accomplishment
service interruption
availability = MTTF / (MTTF + MTTR)
MTTR: mean time to repair
MTTF: mean time between failure
2014147526
현채연